Automated machine learning predicts male infertility based on the Johnsen score

Infertility affects women and men equally. In male infertility, azoospermia (a disease with no semen in semen) is a major problem that prevents a couple from having a child.

For the treatment of patients with azoospermia, testicular sperm extraction (TESE) is required to obtain mature sperm. Upon examination, histological samples are typically given a rating called a Johnsen rating, on a scale of 1 to 10, based on the histopathological features of the testicle.

The Johnsen Score has been widely used in urology since it was first reported 50 years ago. However, histopathological assessment of the testicle is not an easy task and takes a long time due to the complexity of testicular tissue resulting from the numerous highly specialized steps in spermatogenesis. “

Dr. Hideyuki Kobayashi, Associate Professor, Department of Urology, Toho University Medical Faculty

“Our goal was to simplify this time-consuming diagnostic step by leveraging AI technology. To do this, we chose Google’s vision for automated machine learning (AutoML), which requires no programming to create an AI model for individual patient records AutoML Vision allows non-programming clinicians to use deep learning to create their own models without the assistance of data scientists, “said Dr. Hideyuki Kobayashi, Associate Professor of Urology at Toho University School of Medicine.

“The model we created can classify histological images of the testicle without the help of pathologists. I hope our approach will enable clinicians in all areas of medicine to create AI-based models that can be used in their daily clinical practice,” he said.

To simplify the use of Johnsen scores in clinical practice, Dr. Kobayashi four designations: Johnsen score 1-3, 4-5, 6-7 and 8-10. He and his co-researchers obtained a data set of 7155 images at 400x magnification. All images were uploaded to the Google Cloud AutoML Vision platform. For the magnified X400 image dataset, the algorithm average accuracy (positive predictive value) was 82.6%, the accuracy was 80.31%, and the recall was 60.96%.

AI has become popular and is used in all areas of medicine. However, the use of AI by clinicians in hospitals is still hampered by the need for help from data scientists in properly using AI.

“The cloud-based machine learning framework we’ve been using is for everyone. It can become such a powerful tool in medicine that hospital doctors will easily be using AI-based medical image classifiers in the near future, as well as they now use Microsoft PowerPoint or Excel, “said Dr. Kobayashi. He added, “The hardest part was taking pictures of the testicular pathology and it was very time consuming. Two colleagues worked very hard to get all of the images used in the study. I really appreciate their dedicated efforts.”

Dr. Kobayashi’s group has described the development of an AI-based algorithm for evaluating Johnsen scores, in which original images (X400) were combined to achieve a high level of accuracy. This is the first report of an algorithm that can predict Johnsen scores without relying on pathologists and data science experts.


Journal reference:

Ito, Y. et al. (2021) A Method of Using Automated Machine Learning to Histopathological Classification of Testes Based on Johnsen Scores. Scientific reports.

Comments are closed.